Date: Wed, 7 Feb 2024 15:48:18 +0100
From: Krzysztof Kozlowski <krzysztof.kozlowski@...aro.org>
To: Peter Griffin <peter.griffin@...aro.org>
Cc: arnd@...db.de, linux@...ck-us.net, wim@...ux-watchdog.org,
 alim.akhtar@...sung.com, jaewon02.kim@...sung.com,
 semen.protsenko@...aro.org, kernel-team@...roid.com,
 tudor.ambarus@...aro.org, andre.draszik@...aro.org, saravanak@...gle.com,
 willmcvicker@...gle.com, linux-fsd@...la.com,
 linux-watchdog@...r.kernel.org, devicetree@...r.kernel.org,
 linux-kernel@...r.kernel.org, linux-arm-kernel@...ts.infradead.org,
 linux-samsung-soc@...r.kernel.org
Subject: Re: [PATCH v2 1/2] soc: samsung: exynos-pmu: Add regmap support for
 SoCs that protect PMU regs

On 07/02/2024 12:42, Peter Griffin wrote:
>>>>>  #include <linux/soc/samsung/exynos-regs-pmu.h>
>>>>>  #include <linux/soc/samsung/exynos-pmu.h>
>>>>>
>>>>>  #include "exynos-pmu.h"
>>>>>
>>>>> +static struct platform_driver exynos_pmu_driver;
>>>>
>>>> I don't understand why do you need it. You can have only one
>>>> pmu_context. The moment you probe second one, previous becomes invalid.
>>>>
>>>> I guess you want to parse phandle and check if just in case if it points
>>>> to the right device, but still the original code is not ready for two
>>>> PMU devices. I say either this problem should be solved entirely,
>>>> allowing two devices, or just compare device node from phandle with
>>>> device node of exynos_pmu_context->dev and return -EINVAL on mismatches.
>>>
>>> Apologies I didn't answer your original question. This wasn't about
>>> having partial support for multiple pmu devices. It is being used by
>>> driver_find_device_by_of_node() in exynos_get_pmu_regmap_by_phandle()
>>> to determine that the exynos-pmu device has probed and therefore a
>>> pmu_context exists and a regmap has been created and can be returned
>>> to the caller (as opposed to doing a -EPROBE_DEFER).
>>>
>>> Is there some better/other API you recommend for this purpose? Just
>>> checking pmu_context directly seems racy, so I don't think we should
>>> do that.
>>
>> Hm, I don't quite get why you cannot use of_find_device_by_node()?
> 
> of_find_device_by_node() returns a platform_device, even if the driver
> hasn't probed. Whereas driver_find_device_by_of_node() iterates
> devices bound to a driver.
> 
> If using of_find_device_by_node() API I could check the result of
> platform_get_drvdata(), and -EPROBE_DEFER if NULL (that pattern seems
> to be used by a few drivers). But that AFAIK only guarantees you
> reached the platform_set_drvdata() call in your driver probe()
> function, not that it has completed.

All drivers, except two, use of_find_device_by_node(), so basically you
claim they are all broken. If that's true, the core API and these
drivers should be fixed, instead of implementing here entirely different
pattern.

of_find_device_by_node() goes via platform_bus_type->sp->klist_devices
and devices are added to the list in device_add() after
bus_probe_device(dev), regardless of its success. Therefore after
successful first probe, you will have the same result.


> 
> IMHO the drivers using driver_find_device_by_of_node() for probe
> deferral are doing it more robustly than those using
> of_find_device_by_node()  and checking if platform_get_drvdata() is
> NULL.

Some are checking dev->driver, but this also looks buggy, because it is
called before actual drv->probe().

OK, let's go with this method. I dislike the difference from everyone
else, but it seems everyone else is doing it wrong. :(

Best regards,
Krzysztof

Powered by blists - more mailing lists